Measuring Inflation Contents

Chapter 2: Criticism of the Retail Prices Index

55.There have been several reviews of RPI in recent years, both by the statistical authorities and independent experts. This spate of work was prompted by a change in 2010 to the collection of clothing prices. The result was that RPI lost its status as a National Statistic in 2013 and is now treated as a ‘legacy measure’ by the statistical authorities. This means that the statistical authorities will carry out no further work to improve the methodology of RPI, beyond routine improvements such as updating the inflation basket and expenditure weights.46

56.This chapter examines the recent criticism of RPI and considers whether its status as a legacy measure is justified and tenable.

Recent issues with RPI

57.As explained in the previous chapter, the CPI and RPI use different formulas to aggregate prices where expenditure weights are not available. This leads to a difference in their estimates of inflation. This difference is referred to as the ‘formula effect’. Figure 2 shows how much the Office for National Statistics has calculated that RPI has been higher than CPI since 2005 due to this formula effect.

Figure 2: Difference between 12 month changes in RPI and CPI due to the formula effect, January 2005 to November 2018 (%)47

Line graph showing difference between 12 month changes in RPI and CPI due to the formula effect, January 2005 to September 2018 (%)

Source: Office for National Statistics, ‘Difference between CPI and RPI due to formula effect 2015=100’ (19 December 2018): https://www.ons.gov.uk/economy/inflationandpriceindices/timeseries/dra9/mm23

58.As Figure 2 shows, there was a large sudden increase in the formula effect in 2010. In December 2009, RPI was estimated to be 0.54 percentage points higher than CPI as a result of the formula effect; by December 2010, it was estimated to be 0.86 percentage points higher.48 What happened in 2010?

Effect of the changes to clothing price collection

59.The Office for National Statistics said that the increase was mainly due to methodological changes to the measurement of clothing prices that were implemented from January 2010. Changes were made due to difficulties in collecting comparable prices for clothing items throughout the year.49 These changes included an increased sample size from relaxing rules on the comparability of different clothing styles,50 and the collection of prices from items that are on sale in January. The changes were made as measured clothing inflation prior to 2010 was considered to be too low.

60.The effect of this change can be seen when the RPI’s estimates of the annual price change in ‘women’s outerwear’ (one of the classes of goods included in the RPI basket) are compared before and after 2010, as shown in Figure 3.

Figure 3: RPI’s estimate of the annual price change of women’s outerwear, 1987 to 2017 (%)

Line graph showing RPI's estimate of the annual price change of women's outerwear, 1987 to 2017

Source: Office for National Statistics, ‘Consumer price inflation time series’ (19 December 2018): https://www.ons.gov.uk/economy/inflationandpriceindices/datasets/consumerpriceindices

From 1987 to 2009, the average annual price change in women’s outerwear as measured by the RPI was a 2.5 per cent decrease; from 2010 to 2017, the average annual price change was an 11.1 per cent increase. It is implausible that such a large swing is due to actual price rises since 2010.

61.The CPI was also affected by the change (both indices use the same price quotes for clothing), although the effect on RPI was greater. Figure 4 compares CPI and RPI estimates of the annual price change for all clothing items.51

Figure 4: Comparison between the CPI’s and RPI’s estimate of the annual price change of clothing items, 2005 to 2017 (%, CPI class: ‘clothing and footwear goods’; RPI class: ‘clothing and footwear’)

Line graph showing comparison between the CPI's and RPI's estimate of the annual price change of clothing items, 2005 to 2017

Source: Office for National Statistics, ‘Consumer price inflation tables’, Tables 30 and 43 (14 November 2018): https://www.ons.gov.uk/economy/inflationandpriceindices/datasets/consumerpriceinflation

62.The Office for National Statistics said that the change to clothing price collection in 2010 was responsible for 0.30 percentage points of the 0.32 increase in the overall formula effect between December 2009 and December 2010 that was highlighted above.52

63.The clothing change had a greater effect on RPI because of its use of the Carli formula (which uses the arithmetic mean) whereas the CPI uses the Jevons formula (which uses the geometric mean) when they are used to calculate the average price change.53 The inclusion of the new data led to greater variability in the ‘price relatives’ (the ratio of the current period price to the base period price). The Office for National Statistics said this meant that the Carli formula returned greater price changes:

“As a general mathematical result, the geometric mean of a given set of values [in this case, the price relatives] is always lower than an arithmetic mean … except when those values are all equal (there is then no difference). The scale of the difference depends on the dispersion of the price relatives (i.e. the ratio of the current period and base period prices); as the variance of the price relatives increases, so does the difference between the geometric mean and arithmetic mean results.

The reason why this is relevant for clothing is that all the improvements to the measurement of clothing have the potential to increase the dispersion of the price relatives.”54

64.Witnesses were critical of the way in which the 2010 change to clothing price collection had been introduced. Jill Leyland, a member of the National Statistician’s advisory panel on consumer prices, said that the change to clothing price collection had been considered to be “a relatively minor technical matter, was not discussed by any advisory committee and, disastrously, was not tested before implementation.”55 Arthur Barnett, a member of the RPI/CPI User Group, said that testing and recovery planning were “particularly important … ONS and UKSA cannot simply assume that problems will not occur when methodology changes are made—there needs to be a plan in place before changes are made to recover in the event of problems.” He said the failure to test the new price collection methodology properly, and to have a recovery plan in place, had resulted in a “collective loss of confidence” amongst the statistical authorities.56

65.The unexpected effect of the change to clothing price collection was the catalyst for a series of assessments of RPI by the statistical authorities, as detailed below. The result was that RPI has been left as a ‘legacy measure’ and no further methodological improvements will be made to it.

Response from the statistical authorities

66.The effect of the clothing price change prompted a programme of work from the Office for National Statistics. They commissioned a report from Professor Erwin Diewert, a leading expert in the index number theory, to provide advice. Professor Diewert’s 2012 report recommended that the Carli index should no longer be used in the RPI and should be replaced with another formula.

67.The then National Statistician, Jil Matheson, issued a public consultation on the future of the RPI following Professor Diewert’s report.57 This put forward four options for improving RPI:

(a)No change;

(b)Replace the Carli formula for clothing only;

(c)Replace the Carli formula for all items;

(d)Change RPI so its formulas align fully with the CPI.58

68.Of the 406 respondents, 322 recommended the first option, with concerns expressed about the effect on index-linked gilts and pensions.59

69.In the National Statistician’s 2013 response to the consultation, she concluded that the Carli formula was the “primary source of the formula effect difference between the RPI and the CPI, and that this formulation does not meet current international standards.”60 She recommended however that the RPI should be maintained in its current form:

“… there is significant value to users in maintaining the continuity of the existing RPI’s long time series without major change, so that it may continue to be used for long-term indexation and for index-linked gilts and bonds in accordance with user expectations.”61

70.The National Statistician also recommended the production of a new additional index—Retail Price Index Jevons (RPIJ)—which would be the RPI but with the Carli formula replaced with the Jevons formula.

71.The UK Statistics Authority accepted these recommendations. It removed National Statistic status from RPI in March 2013. In July 2013 it announced that the RPIJ would become a National Statistic.62 The authority also commissioned a review of consumer price statistics by Paul Johnson.

RPI as a legacy measure

72.Paul Johnson’s review, published in January 2015, described the existing situation:

“By not changing its calculation to make it a more ‘correct’ measure of inflation in the face of clear evidence that the current methodology is flawed, the [UK Statistics] Authority has set a clear precedent. The National Statistician recognised that by stating that the methodology of the RPI would remain unchanged. This means that improvements to the methodology of the CPI and CPIH will not be carried over to the RPI.”63

73.The review recommended that the RPI should be considered a ‘legacy measure’ only; no further changes should be made to it and it should stop being used:

74.The statistical authorities accepted these recommendations from the Johnson review. The UK Statistics Authority told the Committee that the Office for National Statistics considers RPI to be “a legacy measure” and that it publishes it only because it is required to do so by the 2007 Act.65 Paul Johnson however told us that given the continued use of RPI, he had changed his mind since the review and believed changes should be made to it, this is discussed below.66

75.The review also said that RPIJ should be discontinued, given it was not widely used and “seems to cause confusion, with users not clear whether they should move to RPIJ or CPIH.”67 The Office for National Statistics announced it was discontinuing the index in November 2016.68

Current position of the statistical authorities on RPI

76.In March 2018 the Office for National Statistics published a paper on the ‘Shortcomings of the Retail Prices Index as a measure of inflation’.69 The paper described RPI as “a very poor measure of general inflation, at times greatly overestimating and at other times underestimating changes in prices and how these changes are experienced.” It said there were “better measures available” which were “far superior” to RPI.70

77.The paper said that the mathematical properties of the Carli formula increase the growth rate in the RPI by around 0.7 percentage points when compared with the CPI. It also said that the problems with RPI are “many: it is not just the use of the Carli formula that is a concern.” Other than the formula effect, the problems it listed with the RPI were:

78.The current National Statistician, John Pullinger, listed these other problems when he gave evidence to us in July 2018.73 The Deputy National Statistician, Jonathan Athow, told us that “there is quite a fundamental set of issues with RPI.”74

Should RPI be a legacy measure?

79.We heard arguments for and against the position of the statistical authorities on RPI. We here consider the merits of the arguments they have given for treating RPI as a legacy measure and then look at how tenable is the position not to improve RPI.

Shortcomings of the Carli formula

80.Chris Giles, the economics editor of the Financial Times, told us that the Carli index had “known biases”. He said ‘price bounces’—where prices go up but revert back again—exacerbated these biases and this was evident from the effects of the clothing change.75

81.Professor Martin Weale, a former member of the Monetary Policy Committee of the Bank of England, gave an example of the Carli formula’s “upward bias”. It is reproduced in Box 5, with slight modifications.

Box 5: Illustrating upward bias in the Carli formula

In January 2016, there are two shirts which each cost £20.

In January 2017, the price of shirt x rises to £30 whilst the price of shirt y remains at £20.

In January 2018, the price of shirt x drops back to £20 whilst the price of shirt y remains at £20.

Under the Carli formula,76 the average price change from January 2016 to January 2017 would be 25 per cent: (1+1.5)÷2 = 1.25

From January 2017 to January 2018, the average price change would be a 17 per cent reduction: (1+0.66)÷2 = 0.83

When these price change ratios are multiplied together, the increase in price recorded by the Carli index from January 2016 to January 2018 is 4.1 per cent (1.25 × 0.83 = 4.1), even though both shirts cost £20 in January 2016 and £20 in January 2018.

Source: Written evidence from Professor Martin Weale (RPI0017)

82.Dr Ben Broadbent, Deputy Governor of Monetary Policy at the Bank of England, said the problem described in Box 5 was “worse the greater the dispersion of the inflation rates.” He said that what happened with the clothing change in 2010 was that “there was a great deal more dispersion in the inflation rates of the individual items. That is what suddenly pushed [the RPI] up.”77

83.Dr Broadbent said that the failings of the Carli index “have been evident for a century”, citing a pamphlet by the American economist Irving Fisher from 1922.78 Referring to Irving Fisher’s work in a November 2018 letter to the Royal Statistical Society, Dr Broadbent said that “almost a century later there is virtually no other statistical agency in an advanced economy that uses it.”79

84.Robert Hill, Professor of Macroeconomics at the University of Graz, also cited Irving Fisher’s work and concluded that the use of the Carli formula was “indefensible”:

“Irving Fisher warned against using the Carli formula in his 1922 book on index numbers. Carli fails the time reversal test, and suffers from a systematic upward bias. For example, if prices change from periods 1 to 2, but then in period 3 return to their original period 1 levels, a chained Carli index will always find that the price level is higher in period 3 than in period 1 (except in the special case where all prices change by exactly the same proportion from one period to the next).”80

85.When reaching their conclusions, both Professor Diewert’s 2012 report and Paul Johnson’s 2015 review gave weight to the fact that the Carli formula failed this so-called ‘time reversal’ test when coming to their conclusions. Paul Johnson’s review concluded that this property of the Carli formula “is likely to be contributing to the formula effect.”81

86.This feature of the Carli formula was identified by William Stanley Jevons in 1863. Jevons gave the example of an index measuring the price change of cocoa and cloves. If the price of cocoa is doubled and the price of cloves is halved, the Carli formula will say prices have increased by 25 per cent. Jevons said this was logically inconsistent as there has been “no alteration of price whatever.” The geometric mean however will return an answer of no price change. For this reason, Jevons said the geometric rather than arithmetic mean was preferred.82

Challenging criticisms of the Carli formula

87.Some witnesses felt that much of the criticism of the Carli formula was unfounded. The main argument was that the Carli formula had been problematic only when used with the new clothing price data, the implementation of which had been flawed. Otherwise the Carli formula did not lead to substantial upward bias.

88.The RPI/CPI User Group, an independent user group that operates under the auspices of the Royal Statistical Society, said that the problems with the Carli formula has been known about “for decades” and had been “greatly overstated since 2010.”83 They said it was the change to the method for collecting clothing prices, combined with the use of the Carli formula, which was the problem. Arthur Barnett said that the available evidence “does not point to a systematic upward bias. This question is separate from the clothing issue which [relates to] the price collection methodology.”84 The way the clothing change was implemented without proper testing was criticised by witnesses.85

89.The Royal Statistical Society said it was a “fundamental misjudgement” to blame the Carli formula solely for the widening of the formula effect. They said that since 2010 the RPI had “clearly over-estimated” the rise in clothing prices and the CPI clothing index looked more plausible, “but, due to the nature of the Jevons index, it very probably under-estimates it to a certain extent.” They said none of the index formulas could cope properly with an immense variation in prices.86

90.They also said it was “a common misconception” that the Carli index produced the upward bias as described in Box 5:

“In fact, there are two versions of the Carli index—direct and chained. The chained version does indeed suffer from the problem mentioned [above]. However, it is the direct index, which does not behave in this way, which is used in the RPI. Although the RPI is a chain-linked index, the Carli component is not chained. Unfortunately, the Johnson Review does not make this distinction.”87

91.‘Chain-linking’ is the process which enables a continuous price index at the higher class level—for example, potatoes—to be produced. It takes account for the fact that expenditure weights will change from year to year, allowing comparison with previous years.88

92.Dr Mark Courtney, former Head of Economics in the Regulatory Impact Unit, Cabinet Office, said that the process of chaining “might itself cause an upwards or downwards bias, particularly if prices around the chaining point are not smooth but ‘bouncing’”, but there was no evidence of a substantial bias in the RPI:

“This possibility—initially alleged by the ONS as being a weakness of the RPI—has been extensively studied, and where, as in the RPI, chaining takes place annually (rather than monthly) and at an aggregate level, there is no theoretical reason to expect an upward bias, and the best empirical evidence is that chain drift in the RPI is no more than 0.02 percentage points per annum.”89

93.Arthur Barnett said that as RPI does not use a chain-linked Carli index, the transitivity test does not apply. He said that all British consumer price indices failed this test because all indices used the arithmetic mean in chain-linking.90

94.The Royal Statistical Society produced an example of a chained and a direct Carli index which is reproduced in Appendix 4.

95.Paul Johnson’s 2015 review noted that “experts differ on the relative importance of the tests, and an index failing a test does not necessarily mean that index is unsuitable.”91

Other shortcomings of the RPI

96.Paul Johnson’s 2015 review also relied on the other shortcomings of RPI which were identified by the Office for National Statistics: “Issues with the data source of the weights, population coverage and treatment of some goods (like insurance and owner-occupiers housing costs) make the RPI less suitable as a measure of overall inflation.”92

97.As the previous chapter showed, none of these considerations are new. Arthur Barnett asked how the RPIJ—a measure which also had these purported problems—could have been designated a National Statistic in 2013: “therefore five years ago there was no hint that the problem ONS and UKSA perceived with the RPI was other than the Carli formula … This change of mind … would appear to be very recent.”93

98.Jill Leyland said that not everyone agreed that these features of RPI were unsatisfactory. There was a “fundamental difference” between these issues and the clothing issue: “These features were each adopted after long and careful consideration and consultation, as can be seen in the reports of the [Cost of Living] Advisory Committee. In other words there were, in each case, solid grounds for adopting the methodology in question.”94

99.We heard evidence that the Carli formula, as used in the RPI, produces an upward bias. But expert opinion on the shortcomings of the RPI differs.

100.There is however broad agreement that the widening of the range of clothing for which prices were collected has produced price data which, when combined with the Carli formula, have led to a substantial increase in the annual rate of growth of RPI.

101.We are not in a position to reach a conclusion on the question of whether the Carli formula is problematic in areas other than clothing. Given the properties of the Carli formula that may lead to upward bias have long been evident, yet expert opinion still differs, it may be a perpetual debate.

The decision to make no further improvements to RPI

102.There were however witnesses on both sides of the debate who questioned the resolution of the statistical authorities to make no further improvements to RPI. Many witnesses referred to the legal duties of the statistical authorities in producing official statistics. These duties are set out in Box 6.

Box 6: Summary of the legal duties of the UK Statistics Authority when producing official statistics

Under section 7(1) of the Statistics and Registration Service Act 2007 the UK Statistics Authority has the objective of “promoting and safeguarding the production and publication of official statistics that serve the public good.”

Section 7(2) provides that the reference to serving the public good includes, in particular, “informing the public about social and economic matters” and “assisting in the development and evaluation of public policy”.

Section 7(3) provides that the UK Statistics Authority is to promote and safeguard “the quality of official statistics”, “good practice in relation to official statistics” and “the comprehensiveness of official statistics”.

Section 7(4) states that references to the quality of any official statistics includes “their impartiality, accuracy and relevance” and “their coherence with other official statistics”.

Section 21(1) requires the UK Statistics Authority to “compile and maintain the Retail Prices Index” and “publish it every month”.

As discussed in the previous chapter, section 21(2) requires the UK Statistics Authority, “before making any change to the coverage or the basic calculation of the Retail Prices Index”, to “consult the Bank of England as to whether the change constitutes a fundamental change in the index which would be materially detrimental to the interests of the holders of relevant index-linked gilt-edged securities.”

Section 21(3) provides that “If the Bank of England considers that the change constitutes a fundamental change in the index which would be materially detrimental to the interests of the holders of relevant index-linked gilt-edged securities, the Board [of the UK Statistics Authority] may not make the change without the consent of the Chancellor of the Exchequer.”

Source: Statistics and Registration Service Act 2007

103.Chris Giles said that these legal duties meant there was “a fundamental point of principle” that the statistical authorities should produce an “as accurate as possible RPI”.95 Leaving RPI as it is was “the worst option, because we potentially have an index that is deficient but is central to public life in this country until 2068 … I do not think [this] is tenable in the long term.”96 The RPI/CPI User Group said it believed the statistical authorities were “required under its duty and obligations, to work toward the RPI regaining its [National Statistic] status.”97

104.Simon Briscoe, a consultant on statistical matters, believed that given the legal statutory requirement to promote and safeguard the quality of official statistics, the fact the Office for National Statistics describes RPI as a “poor” measure of inflation implies it is not meeting that statutory duty.98 Unison said that the Office for National Statistics was “failing in its duty to users of inflation measures.”99

105.Geoff Tily, a member of the stakeholder advisory panel on consumer prices (which provides independent advice to the National Statistician), said that “the statistical authorities have been holding the line of the Johnson review ever since it was cast in stone”:

“… at the first meeting of the new advisory panel arrangement we were presented with a copy of the Johnson review almost as if it was the holy writ and we would be implementing it. You cannot get a selection of people together and expect them necessarily to comply, but there has been that kind of ‘groupthink’, if you like.
The momentum behind it has been very much that we are there to implement Johnson.”100

106.Paul Johnson told us that he had changed his mind since his 2015 review. He said his recommendation to make no further improvements to RPI was predicated on it being phased out: “That has not happened very much in the public or the private sector and much more slowly than one might have hoped.”101 He continued:

“you should consider asking [the Office for National Statistics] to correct the RPI rather than stick with its current policy of leaving it as it is … my mind has changed over the past several years because it has become apparent that it is simply not possible for a lot of users of the RPI to move away from it.”102

He said that while it would be “far from perfect” to reverse the clothing price change, “at least we would have something that was consistent over time.”103

Correcting the change to clothing price collection

107.Simon Briscoe said that given that clothing comprised a very small proportion of the overall RPI, it made sense to fix the issue rather than abandon RPI:

“Since 2010, there have been many meetings where people have discussed strappy tops, because it is the most ludicrous example of where the price index has got it wrong. We have to bear in mind that strappy tops are one-thirtieth of one per cent of the RPI. I can think of no other area of life or public policy where if one three-thousandth of something was wrong, we would discard the whole lot. We would simply mend it.”104

108.Jill Leyland said the only change that needed to be made to RPI in the immediate future was to address the clothing problem. A simple solution would be to remove clothing, “or at least those clothing items which swing widely in price” from price indices: “While not an ideal solution it would make the overall index more accurate than at present … and could be used as a temporary measure pending a more permanent solution.”105

Reluctance of the statistical authorities to make the clothing change

109.Jonathan Athow, the Deputy National Statistician, told us there were two reasons why correcting the clothing change was not desirable:

“One is the question whether we could actually achieve it. Could we go back to the 2010 clothing price correction? The clothing market has moved on. Could we try to wind back to the 2010 method? I do not know whether that is possible. I am also reluctant about half-fixing it. It is like having a completely broken down car and putting new tyres on it … I would not describe an RPI rewound to 2010 as anywhere near good … we could end up with new problems from price collection.”106

110.We asked the National Statistician, John Pullinger, why the UK Statistics Authority would not at least correct the clothing problem. He replied that “If were to make [that] change, I would want to deal with all the things that I think need to be changed.”107

111.Sir David Norgrove, Chairman of the UK Statistics Authority, told us that following the procedure in section 21 of the 2007 Act, as set out in Box 6, the Chancellor would refuse to make such a change:

“I could go through the stately dance of going to the Bank of England and saying that we want to change it. It would then say, “It’s a material detriment. Go to the Chancellor”. The Chancellor would then refuse to change it. We may need to do that, but I would much rather work with the people concerned to see whether we can come up with a solution.”108

112.Liz Truss MP, the Chief Secretary to the Treasury, told us that “it is difficult for the Chancellor to say yes or no to a request that has not been submitted … I am suggesting it is the role of the ONS to put forward that proposal.”109 Philip Hammond MP, the Chancellor of the Exchequer, told us that the Office for National Statistics, “should it wish to do so, would find its way to my office relatively easily and I would be very happy to hear from them if they wished to talk to me.”110

Detriment to holders of index-linked gilts

113.If the clothing price change were to be reversed, this would be expected to lower RPI inflation (the change in 2010 was believed to have increased RPI by around 0.3 percentage points111). This would mean that the interest payments on index-linked gilts, which are uprated in line with RPI, would be expected to be lower. If the Bank of England deemed the clothing change to be “fundamental” and “materially detrimental” to index-linked gilt holders, the Chancellor of the Exchequer would have to approve the change.

114.The National Statistician suggested that this provision was behind the reluctance to correct the clothing change:

“We made the change in 2010 and it has had a larger set of consequences than was anticipated then. In the light of that experience, any change is now subject to this consideration that is baked into the legislation about a fundamental change that may or may not be materially detrimental to one party …  The changes we are talking about here, it seems to me, would be fundamental, although that is a matter for the Governor to consider.”112

115.He explained how he perceived his legal duty to produce statistics that serve the public good:

“The RPI represents a series of accumulated expectations of all kinds of parties. Any change that we were to make to it would result in some getting a windfall gain and some getting a windfall loss …

… as I interpret it, my statutory duty is to produce the statistics that serve the public good as the public good is perceived by me. I am hearing three different types of request to me, each of which I am currently seeking to fulfil …

… We have these three different sets of [requests], the third of which is the position that has arisen since 2010, where it is very hard to move from the RPI as it currently is because of these expectations, one way or the other. I am being urged. I do not see this as political pressure. I see this simply as a user requirement. Simply because these contracts [index-linked gilts] are denominated in RPI with these characteristics, and people are in them with a set of expectations around them, it is tricky to fix it, either in part or in whole.”113

116.Given its widespread use, it is surprising that the UK Statistics Authority is treating RPI as a ‘legacy measure’. The programme of periodic methodological improvements should be resumed.

117.We are unconvinced by the National Statistician’s suggestion that in publishing statistics that serve the public good, the interests of those who may be affected negatively by any change should be taken into account. It is not clear from section 7 of the Statistics and Registration Service Act 2007 that this is a relevant consideration for the statistical authorities to be taking into account when they are producing and publishing statistics.

118.What is clear from section 7 is that the UK Statistics Authority has to promote and safeguard the quality of official statistics, which includes their impartiality, accuracy and relevance, and coherence with other statistics. In publishing an index which it admits is flawed but refuses to fix, the Authority could be accused of failing in its statutory duties.

119.We believe section 7 requires the Authority to attempt to fix the issue with clothing prices. Section 21 may require the Authority to consult the Bank of England over the change and obtain the consent of the Chancellor of the Exchequer, however this provision cannot be cited as a reason for not requesting the change in the first place.

120.If the Authority requests the change, the Chancellor of the Exchequer should consent to it. It is untenable for an official statistic, that is used widely, to continue to be published with flaws that are admitted openly.


46 Along with changes made to CPI and CPIH that need also to be made to RPI due to the common elements of compilation.

47 January 2005 is the earliest date for which the Office for National Statistics has calculated the formula effect.

48 Office for National Statistics, ‘CPI and RPI: increased impact of the formula effect in 2010’ (2011): https://webarchive.nationalarchives.gov.uk/20160106041638/http://www.ons.gov.uk/ons/guide-method/user-guidance/prices/cpi-and-rpi/cpi-and-rpi--increased-impact-of-the-formula-effect-in-2010.pdf [accessed 21 December 2018]

49 Robert O’Neill et al (2017), Inflation: History and Measurement, Palgrave Macmillan, p 292

50 Previously, small changes to the composition or style of a garment were treated as a quality change, and therefore the price in the current period could not be compared to the base period.

51 This was the highest level at which comparable data on clothing was available to compare CPI and RPI.

52 Office for National Statistics, ‘CPI and RPI: increased impact of the formula effect in 2010’ (2010): https://webarchive.nationalarchives.gov.uk/20160106041638/http://www.ons.gov.uk/ons/guide-method/user-guidance/prices/cpi-and-rpi/cpi-and-rpi--increased-impact-of-the-formula-effect-in-2010.pdf [accessed 21 December 2018]

53 See para 47 for a description of the difference between the arithmetic mean and geometric mean.

54 Office for National Statistics, ‘CPI and RPI: increased impact of the formula effect in 2010’ (2011): https://webarchive.nationalarchives.gov.uk/20160106041638/http://www.ons.gov.uk/ons/guide-method/user-guidance/prices/cpi-and-rpi/cpi-and-rpi--increased-impact-of-the-formula-effect-in-2010.pdf [accessed 21 December 2018]

55 Written evidence from Jill Leyland (RPI0038)

56 Written evidence from Arthur Barnett (RPI0015)

57 Office for National Statistics, ‘National Statistician’s consultation on options for improving the Retail Prices Index’ (October 2012): https://www.ons.gov.uk/ons/about-ons/user-engagement/consultations-and-surveys/national-statistician-s-consultation-on-options-for-improving-the-retail-prices-index/options-for-improving-rpi-consultation-document.pdf [accessed 21 December 2018]

58 Ibid.

59 Only 20 responses supported one of the options for change, 54 did not express a preference.

60 Office for National Statistics, ‘Response to the National Statistician’s consultation on options for improving the Retail Prices Index’ (February 2013): https://www.ons.gov.uk/ons/about-ons/get-involved/consultations/archived-consultations/2012/national-statistician-s-consultation-on-options-for-improving-the-retail-prices-index/national-statisticians-response.pdf [accessed 21 December 2018]

61 Ibid.

62 UK Statistics Authority, Assessment of compliance with the Code of Practice for Official Statistics, July 2013: https://www.statisticsauthority.gov.uk/archive/assessment/assessment/assessment-reports/assessment-report-257---statistics-on-consumer-price-inflation.pdf [accessed 21 December 2018]

63 UK Statistics Authority, UK Consumer Price Statistics: A Review (January 2015): https://www.statisticsauthority.gov.uk/archive/reports---correspondence/current-reviews/uk-consumer-price-statistics---a-review.pdf [accessed 21 December 2018]

64 Ibid.

65 Written evidence from UK Statistics Authority (RPI0026)

66 See para 106.

67 UK Statistics Authority, UK Consumer Price Statistics: A Review (January 2015): https://www.statisticsauthority.gov.uk/archive/reports---correspondence/current-reviews/uk-consumer-price-statistics---a-review.pdf. [accessed 21 December 2018] The review also said that apart from the use of the Carli formula, the RPIJ suffered from all of the other problems of RPI, discussed below at para 96.

68 Office for National Statistics, ‘Clarification of publication arrangements for the Retail Prices Index and related indices: November 2016’ (10 November 2016): https://www.ons.gov.uk/economy/inflationandpriceindices/articles/clarificationofpublicationarrangementsfortheretailpricesindexandrelatedindices/november2016 [accessed 21 December 2018]

69 Office for National Statistics, ‘Shortcomings of the Retail Prices Index as a measure of inflation’ (8 March 2018): https://www.ons.gov.uk/economy/inflationandpriceindices/articles/shortcomingsoftheretailpricesindexasameasureofinflation/2018–03-08 [accessed 21 December 2018]

70 Ibid. See para 42 for the differences in population coverage between the CPI and RPI.

71 We note that house prices are paid only by those buying a house and interest rates do not affect most homeowners (as the majority of homeowners do not have a mortgage).

72 Ibid. The Living Costs and Food Survey is the successor to the Family Expenditure Survey that was instigated when the RPI began officially in the 1950s: see para 29.

73 Q 53 (John Pullinger)

74 Q 14 (Jonathan Athow)

75 Q 10 (Chris Giles)

76 See Box 4 for an explanation of how the Carli formula works.

77 Q 65 (Dr Ben Broadbent)

78 Q 66 (Dr Ben Broadbent)

79 Letter from Dr Ben Broadbent to Stephen Penneck (8 November 2018): https://www.rss.org.uk/Images/PDF/influencing-change/2018/response-stephen-penneck-8-11-18.pdf [accessed 21 December 2018]. Dr Broadbent quoted from Irving Fisher’s work in the letter: “Fisher was pretty clear in his criticisms of Carli in 1922: “if this book has no other effect than to lead to the total abandonment of the simple arithmetic type of index number it will have served a useful purpose.””

80 Written evidence from Professor Robert Hill (RPI0010)

81 UK Statistics Authority, UK Consumer Price Statistics: A Review (January 2015): https://www.statisticsauthority.gov.uk/archive/reports---correspondence/current-reviews/uk-consumer-price-statistics---a-review.pdf [accessed 21 December 2018]

82 In year 1, the price of cocoa is 1.00 and the price of cloves is 1.00. In year 2, the price of cocoa is 2.00 and the price of cloves is 0.50. Under the Carli formula, the average price change would be (2 + 0.5) / 2 = 1.25, a 25 per cent increase. Under the Jevons formula, the average price change would be (2 x 0.5)1/2 = 1.00, no price change.

83 Written evidence from RPI/CPI User Group (RPI0028)

84 Written evidence from Arthur Barnett (RPI0015)

85 Written evidence from Dr Mark Courtney (RPI0011) and Jill Leyland (RPI0038)

86 Written evidence from Royal Statistical Society (RPI0001)

87 Ibid. They provided an example of a chained and a direct Carli index which is reproduced in Appendix 4.

88 UK Statistics Authority, UK Consumer Price Statistics: A Review (January 2015): https://www.statisticsauthority.gov.uk/archive/reports---correspondence/current-reviews/uk-consumer-price-statistics---a-review.pdf [accessed 21 December 2018]. The weights used in the calculation of RPI have been updated annually since 1962.

89 Written evidence from Dr Mark Courtney (RPI0011)

90 Written evidence from Arthur Barnett (RPI0015)

91 UK Statistics Authority, UK Consumer Price Statistics: A Review (January 2015): https://www.statisticsauthority.gov.uk/archive/reports---correspondence/current-reviews/uk-consumer-price-statistics---a-review.pdf [accessed 21 December 2018]

92 Ibid.

93 Written evidence from Arthur Barnett (RPI0040)

94 Written evidence from Jill Leyland (RPI0038)

95 Q 7 (Chris Giles)

96 Q 6 (Chris Giles). An index-linked gilt was issued in 2013 that matures in 2068, see Chapter 3.

97 Written evidence from RPI/CPI User Group (RPI0028)

98 Q 23 (Simon Briscoe)

99 Written evidence from Unison (RPI0004)

100 Q 30 (Geoff Tily)

101 Q 6 (Paul Johnson)

102 Q 11 (Paul Johnson)

103 Q 6 (Paul Johnson)

104 Q 23 (Simon Briscoe)

105 Written evidence from Jill Leyland (RPI0038)

106 Q 20 (Jonathan Athow)

107 Q 53 (John Pullinger)

108 Q 15 (Sir David Norgrove)

109 Q 33 (Elizabeth Truss MP)

110 Oral evidence taken on 11 September 2018 (Session 2017–19), Q 11 (Philip Hammond MP)

111 See para 58.

112 QQ 53–54 (John Pullinger)

113 Q 50 (John Pullinger)




© Parliamentary copyright 2019